Linking Pore Network Structure to Preferential Pathway Development: Differential Evolution & Shortest Path Analysis
Abstract
Keywords: Shortest Path Analysis, Differential Evolution, Porous Media Structure, Percolation Theory, Anomalous Transport
Previous efforts to characterize preferential flow have largely focused on continuum-scale system descriptors that do not accurately capture the spatial arrangement of the flow. However, it is recognized that large-scale flow observations are the collection of pore-scale interactions between the medium's pore spaces and the flowing fluid. To address this, our study identifies pore-structural attributes that determine overall velocity distribution and spatial arrangement of preferential flow paths in heterogeneous porous media. Two-dimensional samples of polydispersed media producing varying levels of preferential flow were used. Flow fields were resolved in the samples with direct numerical simulations, while pore-structure was measured with pore-network analysis. Borrowing concepts from percolation theory, the percolating (preferential) pathway and stagnant flow regions were identified in each sample by setting velocity thresholds. Simultaneously, the pore-network was used to predict the location of the percolating pathway through shortest path analysis of the network as weighted by different structural attributes. Overlap between the percolating pathway and the predicted shortest path determines the accuracy of the weighting scheme tested. A differential evolution algorithm was employed to determine the "fittest" weighting scheme possible by evolving a defined objective function. This algorithm continuously changes the weight of the structural attribute candidates to make it more or less impactful to the overall weighting scheme. Then, it compares the performance of each weighting scheme generation until convergence to an acceptable level of fitness is achieved. Anticipated results would generate one general weighting scheme that will be able to predict the location of high velocity regions in samples of heterogeneous porous media. The results of this work could be used for fast characterization of porous media heterogeneity, which in turn can be used to predict the propensity of preferential flow and velocity distribution based on structural information alone.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.H31K1862K
- Keywords:
-
- 1805 Computational hydrology;
- HYDROLOGY